[WIP][NV] add minimaxm2.5_fp4_b300_trt.sh

[hshrivastava-droid]

Note

Low Risk
Benchmark-only YAML, shell script, and changelog plus a B300 squash path tweak; no application auth, data, or serving logic changes.

Overview
Adds MiniMax-M2.5 NVFP4 on B300 to the benchmark matrix via a new minimaxm2.5-fp4-b300-trt entry in nvidia-master.yaml (TensorRT-LLM 1.3.0rc18, nvidia/MiniMax-M2.5-NVFP4) with fixed-seq-len sweeps for 1k/1k and 8k/1k across TP/EP and DP-attention vs non-DP search spaces.

Introduces benchmarks/single_node/fixed_seq_len/minimaxm2.5_fp4_b300_trt.sh, which generates a TRT-LLM extra config (CUDA graphs, MoE/NVFP4 backends, optional attention DP), runs trtllm-serve under mpirun, and drives the standard serving benchmark (plus optional lm-eval when RUN_EVAL=true). Documents the config in perf-changelog.yaml.

B300 NV launcher: updates the cluster comment and changes container squash image path from /data/home/sa-shared/gharunners/squash/ to /data/squash/.

^{Reviewed by Cursor Bugbot for commit 9c5522e. Bugbot is set up for automated code reviews on this repo. Configure here.}

[github-actions]

Thanks for the contribution! For vLLM & SGLang, please ensure that your recipes is similar to the official vLLM recipes and/or the SGLang cookbook

• https://docs.sglang.io/cookbook/intro
• https://recipes.vllm.ai/

If it is not, please create a PR first before we can merge your single node PR into the master branch. Let's ensure that the documentation is first class such that the entire ML community can benefit from your hard work! Thank you

• https://github.com/vllm-project/recipes
• https://github.com/sgl-project/sglang/tree/main/docs_new

PR authors are responsible for ensuring that after merging, all GitHub Action jobs fully pass. A lot of the time, failures are just flakes and simply re-running the failed jobs will fix it. If re-running failed jobs is attempted, PR authors are responsible for ensuring it passes. See GitHub's docs on re-running failed jobs: https://docs.github.com/en/actions/how-tos/manage-workflow-runs/re-run-workflows-and-jobs#re-running-failed-jobs-in-a-workflow

As a rule of thumb, generally, PR authors should request a review & get a PR approval from the respective companies' CODEOWNERS before requesting a review from core maintainers.

If additional help is needed, PR authors can reach out to core maintainers over Slack.

[cursor]

Cursor Bugbot has reviewed your changes and found 1 potential issue.

^{❌ Bugbot Autofix is OFF. To automatically fix reported issues with cloud agents, enable autofix in the Cursor dashboard.}

^{Reviewed by Cursor Bugbot for commit 9c5522e. Configure here.}

[cursor]

Missing eval-only context setup

Medium Severity

The script writes max_seq_len from MAX_MODEL_LEN into the TRT-LLM config and starts the server without an EVAL_ONLY branch. Eval-only jobs (supported by the benchmark workflow) never call setup_eval_context, so the server can keep a throughput-tuned context cap and fail or truncate lm-eval runs.

 

^{Reviewed by Cursor Bugbot for commit 9c5522e. Configure here.}

[github-actions]

see unofficial run visualizer at https://inferencex.semianalysis.com/inference?unofficialRun=27375346990
see unofficial run visualizer at https://inferencex.semianalysis.com/evaluation?unofficialRun=27375346990